TL;DR. This document presents a series of experiments exploring latent side task capability in large language models. We adapt Ryan’s filler token experiment into a more AI Control-like setup with main task and side task and find that Claude Opus 4.5 can solve harder arithmetic problems latently when it has...
Authors: Yueh-Han Chen, Robert McCarthy, Bruce W. Lee, He He, Ian Kivlichan, Bowen Baker, Micah Carroll, Tomek Korbak In collaboration with OpenAI TL;DR: Chain-of-thought (CoT) monitoring can detect misbehavior in reasoning models, but only if models cannot control what they verbalize. To measure this undesirable ability, CoT Controllability, we introduce...
This writing is organized in three sections: * Section 0 - A short overview of AI Control * Section 1 - The structure of an AI Control evaluation * Section 2 - AI Control, Next Steps? 0. A short overview of AI Control AI Control is a framework for extracting...
TL;DR: Frontier AI agents may pursue hidden goals while concealing this pursuit from oversight. Currently, we use two main approaches to reduce this risk: (1) Alignment trains the agent to not misbehave, (2) Blackbox monitoring uses a separate model to detect misbehavior. We study a third approach—self-incrimination—which trains agents to...
Introduction My collaborators and I wrote a paper titled "Distillation Robustifies Unlearning" a few months ago. For a quick summary, you can check out my mentor’s Twitter thread. Our team will be presenting posters at NeurIPS and the San Diego Alignment Workshop, so feel free to find me if you...
Current “unlearning” methods only suppress capabilities instead of truly unlearning the capabilities. But if you distill an unlearned model into a randomly initialized model, the resulting network is actually robust to relearning. We show why this works, how well it works, and how to trade off compute for robustness. Unlearn-and-Distill...